Introduction: spatial effects and the role of configuration of data

Abstract

Spatial data in regional economic analysis have two distinctive features. First of all they cannot be thought as randomly generated from the classical urn models; rather they are dependent in that “the value of, say, prosperity in one region gives statistical information about the likely value of adjacent areas” (Unwin and Hepple, 1974). This is usually referred to as the spatial autocorrelation problem (Cliff and Ord, 1981). Secondly they are constituted by aggregation of the characteristics of individuals within portions of space. The population in a country is the sum of the individuals living in that country, the total income of a region is the sum of the income of the population in that region: the per-capita consumption of an area is the mean of the individual consumption of that area, and so on. However the borders of the zones in which a study area is divided are not just divinely given; rather there is a very large number of different ways in which the individuals can be aggregated to form areal data. Analysis is often made more complex by the common situation faced in geographical investigation that the variable of interest is recorded on a system of irregular collecting areas. Thus the study area is divided into territorial units which are, in general, different in size and shape and which connect to one another in an arbitrary and irregular way. The problem of the arbitrariness and irregularity of spatial units will be referred to as the modifiable areal unit problem following Openshaw and Taylor (1979).

Everybody knows that geography is about maps

(Unwin, 1981; p.1)

The fundamental notion in statistical theory is that of group or aggregate

(Kendall & Stuart, 1976; p.1)